A Closed-Loop Approach To Efficient And Stable Supply-Chain Coordination In Complex Stochastic Manufacturing Systems

Modem manufacturing productivity and competitiveness is undoubtedly time-based [2]. As a result, the design of production facilities (Product Cells) and their operation (Just In Time, Zero In Process, Lean etc. [3,4,5,6]) aim at reducing lead times and inventories of either work in process or finished goods. Moreover, the design of manufacturing processes and in general the choice of manufacturing technology (automation, flexible computer controlled machines for single part as opposed to batch processing; selection of type of material, raw material, and intermediate goods used in production for delayed differentiation [46,48]) are significantly, if not primarily, motivated by lead time reduction objectives. Finally, information technology is looked upon as a means to reduce inventories needed to provide a desired service level through supplier-consumer information sharing. For example, knowledge of a retailer's inventory replenishment policy can help its supplier to manage its own inventory better [46].Significant productivity gains in general and in lead time reduction in particular have indeed been made so far by adopting the Japanese Model (simplifying the production process, introducing visual factory standards [5]), by utilizing advanced automation and information technology, and, last but not least, by exploiting the increased set of intermediate goods choices made possible through the expansion of trade [47]. However, for all the advances in decreasing lead times at individual links of the supply chain, potential improvements in global supply chain lead times have not been captured. In particular, when an individual link of the manufacturing supply chain is flexible enough to respond to production requirement fluctuations modulating its own lead time accordingly, the rest of the supply chain can not benefit fully from this flexibility as long as it has no information of the individual link's lead time dynamics. In fact, present practice is for the manufacturing supply chain to plan using a worst case or maximal lead time assumption about each individual link. The assumption that lead times are constant - at best revised periodically to a different constant value - is the basic premise of widely used Material Requirement Planning (MRP) production scheduling algorithms [18]. The MRP practice is a serious impediment to further productivity gains. Indeed, significant gains are possible when lead time dynamics of individual supply chain links are accounted for in the overall manufacturing supply chain coordination through synergistic decentralized production planning. Better production planning algorithms that exploit information sharing are capable of taking the industry to the next big step in productivity gains. This paper argues that time scale decomposition for enhanced information sharing and better deterministic fluid model approximations capturing the essence of the underlying stochastic dynamics can indeed provide these productivity gains.